Cyclic Sensor Trace Analysis for Multi-Run Process Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods face challenges in efficiently analyzing sensor data from multi-run processing procedures with cyclic and looped operations, leading to difficulties in identifying anomalies and performing corrective actions in a timely manner, resulting in sub-optimal product quality and increased downtime.
Innovation Solution
A method that separates time trace sensor data into cycle data sets associated with repeated operations, processes these data sets to generate summary data using statistical and machine learning methods, and provides alerts to users for potential issues, enabling proactive corrective actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to analyze sensor data from multi-run processing procedures, then the analysis process becomes complex and time-consuming, but the ability to identify anomalies in a timely manner deteriorates
Solution Approach 1:
The patent segments the complex sensor data from multi-run processing procedures into individual cycle data sets, each representing a specific cyclic operation. This segmentation allows for targeted analysis of each cycle independently, making it easier to identify anomalies without being overwhelmed by the entire data set. The system separates data from different processing runs and cycles, enabling focused statistical analysis on relevant portions only.
Solution Approach 2:
The patent extracts key features and summary statistics from the segmented cycle data sets, isolating the most relevant information for anomaly detection. By taking out only the essential characteristics from each cycle and comparing them against expected ranges, the system achieves rapid anomaly identification without analyzing every detail of the raw sensor data, thus reducing analysis time while maintaining detection accuracy.
2Measurement precision
If detailed analysis of all sensor data is performed, then measurement precision improves, but productivity decreases due to increased analysis time
Solution Approach 1:
The patent divides the comprehensive sensor data into discrete cycle segments corresponding to specific processing operations. This segmentation enables selective analysis where only relevant cycles are examined in detail, while others are processed through streamlined comparison methods. The system maintains high measurement precision for critical cycles without requiring exhaustive analysis of all data, thereby preserving manufacturing throughput.
Solution Approach 2:
The patent applies partial analysis by focusing computational resources on the most critical cycles and parameters that have the greatest impact on product quality. Rather than performing exhaustive detailed analysis on all sensor data, the system identifies and analyzes only the portions that are most likely to contain anomalies or affect productivity, achieving sufficient measurement precision with reduced analysis time and minimal impact on manufacturing throughput.
3Manufacturing precision
If comprehensive sensor data analysis is implemented, then product quality control improves, but device complexity increases
Solution Approach 1:
The patent segments the data processing system to match the segmented cycle data, creating a modular analysis approach where each segment handles specific cycle types. This modular segmentation reduces overall system complexity by breaking down the complex task of analyzing multi-run processing data into manageable, independent modules that can be processed through standardized procedures, thereby maintaining manufacturing precision without proportionally increasing system complexity.
Solution Approach 2:
The patent implements a universal analysis framework that can handle multiple types of processing cycles and operations through a single set of core procedures. The system uses standardized statistical methods and comparison techniques that apply across different cycle types, reducing the need for specialized complex analysis routines for each specific operation. This multi-functional approach maintains comprehensive product quality control while avoiding the complexity multiplication that would result from having separate specialized systems for each cycle type.
Data Source
AI summary
A method includes receiving time trace sensor data associated with a substrate processing procedure. The substrate processing procedure includes two or more sets of processing conditions. At least a first set of processing conditions and a second set of processing conditions each include one or more operations performed repeatedly. The method further includes separating a first and second portion of the time trace sensor data corresponding to the first and second sets of processing conditions into a first and second plurality of cycle data. The method further includes processing the first plurality of cycle data and the second plurality of cycle data to generate summary data. The method further includes providing an alert to a user. The alert is based on the summary data.